-
Notifications
You must be signed in to change notification settings - Fork 0
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Upstream concepts to Adapt.jl and friends #1
Comments
Looking at this I would be open to take these features on for KA 0.10/1.0 We are getting closer to having KA be used by GPUArrays to provide default implementations |
Thanks @vchuravy! I'll be very happy to "upstream" anything you like in here. |
I think we should still strive to have the basic definition of a compute unit (and of what you can do with it) in a very lightweight package, though. We should be able to do things like "move this data to where that data is" in a very generic fashion in packages without taking on heavy dependencies. While KernelAbstractions isn't a very heavy package, it's not very lightweight either (partially due to StaticArrays :-) ) and packages that don't implement kernels probably won't want to depend on KA. |
Reminder issue. Part of the concepts in HeterogeneousComputing should ideally be upstreamed to Adapt/ArrayInterfaceCore, KernelAbstractions and the GPU packages when ready.
See JuliaGPU/Adapt.jl#64 .
The text was updated successfully, but these errors were encountered: